THE FLATTENING FIRM: EVIDENCE FROM PANEL DATA ON THE
CHANGING NATURE OF CORPORATE HIERARCHIES
Raghuram G. Rajan and Julie Wulf*
Abstract—Using a detailed database of managerial job descriptions,
reporting relationships, and compensation structures in over 300 large
U.S. firms, we find that firm hierarchies are becoming flatter. The number
of positions reporting directly to the CEO has gone up significantly over
time while the number of levels between the division heads and the CEO
has decreased. More of these managers now report directly to the CEO
and more are being appointed officers of the firm, reflecting a delegation
of authority. Moreover, division managers who move closer to the CEO
receive higher pay and greater long-term incentives, suggesting that all
this is not simply a change in organizational charts with no real consequences. Importantly, flattening cannot be characterized simply as centralization or decentralization. We discuss several possible explanations
that may account for some of these changes.
I.
Introduction
E
CONOMIC theorists have long advocated a move away
from seeing the firm as a black box and toward focusing
on its internal organization instead. For example, as Williamson (1981) argues, viewing the firm and its organization
as a “governance” structure rather than simply as a production function would help us understand better the boundaries between the firm and the market.
Work from the 1960s through the early 1980s (see, for
example, Williamson, 1967; Calvo & Wellisz, 1978, 1979;
Rosen, 1982) followed this approach by seeking to explain
the size of firms as a consequence of the limitations on
governance that can be exerted by corporate hierarchies.
Broadly speaking, hierarchies emanate from the need to
supervise workers. Any manager has limited time to supervise employees, so his span of supervision will be limited
(Calvo & Wellisz, 1979; Rosen, 1982), and the number of
layers in the organization will also be limited by the loss of
control across levels (Williamson, 1967; Calvo & Wellisz,
1978). As a bonus, this work also explains why more
talented employees will occupy higher positions in the
hierarchy—because their effort affects more employees (see
Calvo & Wellisz, 1979; Rosen, 1982). Thus at once, two
stylized facts about corporations are explained—that larger
firms pay managers more, and that wages go up as one
moves up the hierarchy.
Received for publication December 7, 2004. Revision accepted for
publication November 22, 2005.
* GSB, University of Chicago; and Wharton School, University of
Pennsylvania, respectively.
We are grateful to Katie Donohue at Hewitt Associates, K. Subramanium, and Talha Noor Muhammed for assistance in data collection. We
would like to thank seminar participants at Wharton, the University of
Chicago, Columbia University, the NBER Organizational Economics
Conference, and the Harvard Business School Strategy Conference, and
especially George Baker, Peter Cappelli, Brian Hall, Brigitte Madrian,
Kevin Murphy, and Joel Podolny. Wulf acknowledges research support
from the Reginald H. Jones Center for Management Policy, Strategy and
Organization at the Wharton School of the University of Pennsylvania,
and Rajan thanks the Center for Research on Securities Prices, the Center
for the Study of the State and the Economy, and the National Science
Foundation for research support.
Since this early work, there has been much more theoretical work trying to explain other aspects of firm hierarchies. Yet we don’t have very many more stylized facts to
discipline the theory, despite the fact that corporations in the
United States have been changing tremendously. Peripheral
businesses have been divested as corporations focus more
on core areas, and peripheral activities have been outsourced [see, for example, the account in Powell (2001)]. At
the same time, large corporations have been merging at a
historically unprecedented rate (see Pryor, 2000). Even
while corporate boundaries are being redrawn, there is some
suggestion that the very nature of employment relationships
is changing (see, for example, Osterman 1996; Holmstrom
& Kaplan, 2001; Rajan & Zingales, 2000).
General Electric’s recent organizational changes illustrate
the type of facts that might be of interest to organizational
theorists. The former chairman of GE Capital, who reported
directly to the CEO, resigned from his position, and the four
business unit heads started reporting directly to the CEO.
Jeffrey Immelt, the CEO of GE, explained the decision thus:
“The reason for doing this is simple—I want more contact
with the financial services teams. . . . With this simplified
structure, the leaders of these four businesses will interact
directly with me, enabling faster decision making and execution.”1 In this example, GE’s organization became flatter:
the CEO’s span of control (or the number of positions
reporting directly to the CEO) increased by 3 (owing to the
loss of the chairman of GE Capital and the gain of four unit
heads: consumer finance, commercial finance, equipment
management, and insurance), and the average number of
reporting levels between the unit heads and the CEO in GE
declined (see figure 1).
Is this pattern of change special to GE, or is it more
systematic? Perhaps a careful documentation of what fundamentally, if anything, has changed in corporate hierarchies will give us a new set of facts to explain, and
hopefully a better way to understand the boundaries between the firm and the market. Certainly, we have learnt a
lot by trying to explain stylized facts about static differences
between firms, and even past changes [for example, Williamson’s (1975, 1985) work on the movement from the
U-form to the M-form of organization]. It is time to add
more facts to the theoretical mill, and to offer preliminary
explanations for them.
We examine how corporate hierarchies have changed in
the recent past. We use a detailed database of job descriptions of top managers, reporting relationships, and compen1 General Electric press release titled “GE Announces Reorganization of
Financial Services; GE Capital to Become Four Separate Businesses,”
July, 26, 2002.
The Review of Economics and Statistics, November 2006, 88(4): 759–773
© 2006 by the President and Fellows of Harvard College and the Massachusetts Institute of Technology
760
THE REVIEW OF ECONOMICS AND STATISTICS
FIGURE 1.—GENERAL ELECTRIC: CHANGE
IN
CEO SPAN
sation structures in over 300 large U.S. firms tracked over a
period of up to 13 years. We focus on the seniormost levels
of the hierarchy: after all, it is the CEO and other members
of senior management who make resource allocation decisions that ultimately determine the firm’s performance (and
most obviously represent the “managers” in the theories).
We document that the flattening of the senior management hierarchy reported in the General Electric example is
widespread in the United States among leading firms in their
sectors.2 Our first finding is that the number of managers
reporting to the CEO has increased steadily over time, from
an average (median) of 4.4 (4) in 1986 to 8.2 (7) in 1998.3
We consider several simple explanations for the increase in
CEO span of control including firm growth, addition of new
positions (such as the chief information officer), and mergers. Taken together, these explanations account for only part
of the trend.
Our second finding is that the depth, which is the number
of positions between the CEO and the lowest managers with
profit center responsibility (division heads), has decreased
by more than 25% over the period.4 Moreover, the number
of division heads reporting directly to the CEO has tripled.
One possible explanation of all this is that the organizational
hierarchy is becoming flatter.
2 Based on a variety of statistical tests, we conclude that our sample is
representative of Fortune 500 firms. We discuss this in detail in section II
B.
3 Others have found, using smaller data sets and focusing on particular
industries, that the manager’s span of control seems to be increasing (see,
for example, Scott, O’Shaughnessy, & Cappelli, 1996), but these studies
typically use an indirect measure of span (the number of managers at one
level divided by the number of managers at the next level) and focus at
levels below the CEO. Our measure of CEO span is potentially more
precise, because we know who reports to the CEO.
4 Baker, Gibbs, and Holmstrom (1994) find that the number of levels is
constant over time for the single firm in their study. Using detailed
personnel records, they infer the number of levels from information about
moves between job titles and consider all levels within the firm. By
contrast, we focus only on the levels between senior management positions, but have a potentially more accurate measure because of information on reporting levels.
OF
CONTROL
AND
ORGANIZATIONAL LEVELS, JULY 2002
Another possible explanation, however, is that fewer but
larger units are being given profit center responsibility. In
other words, it may be that firms have regrouped units into
larger divisions so that division heads have become important enough to report to the CEO. But when we focus only
on divisional manager positions that report over multiple
years (and thus are unlikely to be created or even greatly
affected by organizational restructuring), we find that despite little change in division size, these positions have a
higher probability of reporting to the CEO, as well as a
shorter distance from the CEO on average, over time.
Moreover, more of these positions are getting increased
authority by being denominated “officers” of the firm. So
hierarchies do seem to be getting flatter, even while authority is being delegated down the organization.
One way organizations can become flatter is by eliminating intermediary positions between the CEO and division
heads, as in the GE case. We find evidence of this. For
instance, the chief operating officer (COO), who typically
stood between the CEO and the rest of the firm, is increasingly rare. The number of firms with COOs has decreased
by approximately 20% over the period.
It is always possible that organizational structure is simply a way of conveying status and is otherwise meaningless.
For example, some sociologists argue that informal networks play a much more important role than formal titles
and reporting relationships in determining information
flows and decision-making. To see whether the change in
organizational form has effects outside the minds of managers, we examine how pay changes with organizational
structure. We find both salary plus bonus and long-term
incentives for a divisional manager position increase as it
gets nearer the top, even after correcting for other determinants of pay like the number of employees under the
position’s supervision.
In sum, the CEO seems to be reducing the organizational
distance between him and operational managers such as
division heads. Yet it does not appear he is completely
THE FLATTENING FIRM
taking over the decision-making power or the supervisory
power of the eliminated intermediate layers of management.
Closer divisional managers are getting more authority by
being appointed officers, a fact further corroborated by their
higher pay. Moreover, their greater long-term incentives
suggest that their decisions are being guided by stronger
incentive pay rather than close monitoring. Though the CEO
may be in closer contact with operational managers than
before, he is simply not trying to substitute himself for the
eliminated layers.
Taken together, these findings suggest that corporate
hierarchies are becoming flatter. It is not easy to ascribe the
label “centralization” or “decentralization” to this. On the
one hand, the CEO is getting directly connected deeper
down in the organization, a form of centralization. Increasing span of control suggests he is more directly involved in
decision-making across a greater number of organizational
units. On the other hand, decision-making authority and
incentives are also being pushed further down, a form of
decentralization—or, using the jargon, empowerment.
What could explain the findings? Three possible classes
of explanations are (i) an increase in the competitiveness of
the external environment, forcing the need for a more
streamlined organization, (ii) an improvement in corporate
governance, forcing CEOs to eliminate excessive layers of
managers built up during past empire building, and (iii)
advances in information technology that expand the effective span of control of top managers. Although we will lay
out the rationale for each class of explanations, as well as
possible ways to test them, detailed testing is beyond the
scope of this paper.
We are, of course, not the first to point out that organizations might be becoming flatter. This certainly is conventional wisdom in the business press, and a number of
academic papers have also mentioned it (see, for example,
Powell, 1990; Osterman, 1996; Scott, O’Shaughnessy, &
Cappelli, 1996; Useem, 1996). However, no research we are
aware of systematically quantifies these changes, correlates
them with compensation, and discusses possible explanations of the observed patterns.
The remainder of the paper is outlined as follows. In
section II, we describe the data, in section III we establish
the facts, and in section IV we discuss possible explanations. We conclude in section V.
II.
Data Description
A. The Sample
Empirical work on the organizational structure of firms is
limited. This is primarily due to the lack of detailed information on structures and the difficulty in finding measures
that allow comparisons across firms. As a result, previous
research relies on either detailed data sets of a single firm
[such as the personnel records in Baker et al. (1994)] or less
detailed data on a smaller sample of firms [such as the
761
compensation survey data on 11 insurance firms in Scott et
al. (1996)].5 These studies typically infer the number of
levels in the hierarchy from promotions between positions
or measure the span of control in terms of ratios of the
numbers of employees at different organizational levels. By
contrast, the primary data set used in this study includes a
panel of more than 300 publicly traded U.S. firms over the
years 1986–1998, spanning a number of industries. We use
detailed information on job descriptions, titles, reporting
relationships, and reporting levels of senior and middle
management positions that allow us to characterize organizational structures of firms in a potentially more accurate
way than previous research.
The primary data used in this study are collected from a
confidential compensation survey conducted by Hewitt Associates, a leading human resources consulting firm specializing in executive compensation and benefits.6 The survey is
the largest private compensation survey (as measured by the
number of participating firms) and is comprehensive in that
it collects data on more than 50 senior and middle management positions including both operational positions (such as
the chief operations officer and divisional CEO) and staff
positions (such as the chief financial officer and head of
human resources).7 The survey typically covers all the
positions at the top of the hierarchy and a sample of
positions lower down. An observation in the data set is a
managerial position within a firm in a year. The data for
each position include all components of compensation,
including salary, bonus, restricted stock, stock options, and
other forms of long-term incentives (such as, performance
units). To ensure consistency in matching these positions
across firms, the survey provides benchmark position descriptions and collects additional data for each position,
including the job title, the number of employees under the
position’s jurisdiction, the title of the position that the job
reports to (that is, the position’s boss), and the number of
reporting levels between the position and the board of
directors.
We believe the survey data are accurate, for several
reasons. First, Hewitt personnel are knowledgeable about
survey participants, because they are typically assigned to
specific participants for several years. Furthermore, although the participating firms initially match their positions
to the benchmark positions in the survey, Hewitt personnel
follow up to verify accuracy and spend an additional 8–10
hours on each questionnaire, evaluating the consistency of
responses with public data (such as proxy statements) and
5 There are several early empirical papers on organizational structure
using cross-sectional techniques (for example, Child, 1973; Pugh et al.,
1968).
6 We discuss below (section II B) some possible selection issues associated with this sample.
7 In this study we use a subset of the survey’s benchmark positions:
position descriptions are listed in the appendix.
762
THE REVIEW OF ECONOMICS AND STATISTICS
TABLE 1.—DESCRIPTIVE STATISTICS—WHOLE SAMPLE (UNBALANCED)
AND
BALANCED SAMPLE
A. Firm Characteristics of Sample
Balanced Sample (N ⫽ 51)
Whole Sample (Unbalanced)
Mean
STD
N
Mean
Variable
1986
1998
1998
(firm-yr)
1986
1998
Size (000’s emp.)
Profitability
Age (years)
Number of segments
Inst. shareholders (%)
47.45
0.167
—
2.99
51.6
49.49
0.189
84.8
3.21
61.7
92.27
0.098
40.81
1.85
15.9
3270
3292
3292
2519
2393
85.86
0.162
—
3.29
51.2
73.81
0.201
105.3
3.91
60.9
STD
N
1998
(firm-yr)
106.52
0.093
33.47
1.86
11.2
645
640
640
609
510
B. Industry Characteristics of Firms in Sample
Distribution of Sample by 2-digit SIC Code
Whole Sample
Distribution of Sample by 2-digit SIC Code
Balanced Sample
Whole Sample
Balanced Sample
Industry
(2-digit SIC)
N
(firm-yr)
% of
Sample
N
(firm-yr)
% of
Sample
Industry (2-digit SIC)
N
(firm-yr)
% of
Sample
N
(firm-yr)
% of
Sample
Food (20)
Paper (26)
Chemical (28)
Machinery (35)
Electrical (36)
202
129
467
340
153
6.0
3.9
13.9
10.1
4.6
78
26
169
26
26
12.0
4.0
26.0
4.0
4.0
Transp. equip. (37)
Instrumentation (38)
Communications (48)
Utilities (49)
Other
232
133
161
399
1,134
6.9
4.0
4.8
11.9
33.9
78
26
13
13
195
12.0
4.0
2.0
2.0
30.0
Notes: Whole sample includes all firms in the sample. Balanced sample includes firms that appear in the sample over the 13-year period. Panel A: Profitability is defined as EBITDA/sales. Age is defined as number
of years since founding as listed in the Directory of Corporate Affiliations. Number of segments is that reported in the Business Segment file of Compustat. Institutional shareholders represents the percentage of
shares held by institutions as reported by Spectrum.
across years.8 Participants use the survey results to set pay
levels and design management compensation programs, an
indication that they believe others treat the survey seriously.9
In table 1, we present descriptive statistics for the firms in
the sample. The data set includes more than 300 firms; the
exact number varies over the period, as firms enter and exit
as survey participants. We report statistics on both the whole
sample (unbalanced) and the subset of 51 firms that are
included in the sample for the entire 13-year period (balanced). The firms in the sample are large, publicly traded
U.S. firms that are well established and profitable, with
average size of approximately 47,500 employees, age of 85
years since founding, and return on sales of 19% (see table
1A). The typical firm in the sample is thus a large, mature,
stable firm, not one whose organizational structure is likely
to be in flux. The sample firms span many industrial sectors
of the economy, with some concentration in the food, paper,
chemical, machinery, electrical, transportation equipment,
8 For example, a first-time participating firm reads the position descriptions and is shown examples like the one in figure A1 in order to match
their positions to those covered in the survey.
9 There may be incentives for survey participants to misreport pay data
in their survey responses for positions other than those reported in proxy
statements. However, several facts offset the likelihood of this practice.
First, for Hewitt clients, pay comparisons between the client and survey
averages (excluding the client data) are provided to the board of directors,
making it less likely that clients would misreport their own pay. Second,
these surveys are completed by the firm’s compensation analyst, and it
would require a significant amount of internal coordination among several
managers to intentionally misreport. Finally, the most important measures
in this paper—namely, proxies for span and depth of the hierarchy—aren’t
reported to survey participants and are only used by Hewitt to improve
accuracy in benchmarking positions across firms.
instrumentation, communications, and utilities industries
(table 1B).
B. Sample Representativeness
Clearly, an important question in data sets such as this
one is that of sample selection: whether the firms in the data
set are or are not representative of, employers of similar size
in their industry. The survey participants are typically the
leaders in their sectors; in fact, more than 75% of the firms
in the data set are listed as Fortune 500 firms in at least one
year, and more than 85% are listed as Fortune 1000 firms.
These firms represent a significant fraction of the activity of
publicly traded firms in the United States. Based on all firms
covered in Standard and Poor’s Compustat database over
the period of study, the survey participants represent approximately 33% of employees, 30% of sales, 20% of
assets, and 40% of market value. If we limit the analysis to
manufacturing firms, the Hewitt firms represent 42% of
employees, 38% of sales, 39% of assets, and 52% of market
value.
In general, Hewitt survey participants also participate in
other compensation consulting firm surveys (Hay Associates, Mercer, or Towers Perrin, to name a few) and do so
primarily to receive information about pay practices to use
as a competitive benchmark in evaluating their own compensation programs.10 It is important to note that the sample
10 The value of a compensation survey to a participating firm depends on
how representative it is of firms that the participant competes with in the
executive labor market.
THE FLATTENING FIRM
includes many more firms than Hewitt’s consulting client
base; at least 50% of the firms are survey participants with
no client relationship to Hewitt.11
We evaluate the representativeness of our sample by
comparing key financial measures of our survey participants
with a matched sample from Compustat. We begin by
matching each firm in the Hewitt data set to the Compustat
firm that is closest in sales within its two-digit SIC industry
in the year the firm joins the sample. We then perform
Wilcoxon signed rank tests to compare the Hewitt firms
with the matched firms. Although the firms in the Hewitt
data set are, on average, slightly larger in sales than the
matched sample, we found no statistically significant difference in employment and profitability (return on sales).12
We also found no statistically significant difference in sales
growth, employment growth, or annual changes in profitability for all sample years. In sum, though the Hewitt firms
are larger (measured by sales) on average than the matched
sample, there is little additional evidence that these firms are
not representative of the population of industrial firms that
are leaders in their sectors.13 To sum up, the survey sample
is probably most representative of Fortune 500 firms.
C. Measures of Organizational Structure
Our study focuses on two measures of organizational
structure: the breadth and depth of the hierarchy. Breadth is
represented by the CEO’s span of control (CEO Span) and
is defined as the number of positions reporting to the CEO.
Because we know the title of the position that each position
reports to (the position’s boss), we can determine the positions that report directly to the CEO.14 Our other measure,
depth, represents a vertical dimension of the hierarchy and
11 One concern about sample selection bias is that firms participating in
compensation surveys (Hewitt’s or any others) may be inclined to adopt
more modern compensation practices, including greater incentive pay.
This is certainly possible. However, it is highly unlikely that survey
participants flatten their organizational structure in response to survey
data. As mentioned in an earlier footnote, the information on reporting
levels in the survey is collected to ensure proper benchmarking of
positions across firms, and no information about CEO span of control or
organizational levels of divisional managers is provided to firms in return
for participating.
12 The Hewitt firms are larger in sales than the matched sample of firms
because in a number of the cases, the Hewitt firm is the largest firm in the
industry, thus forcing us to select a matched firm smaller in size.
13 We also calculate financial measures for the sample of Compustat
firms with 10,000 employees or more over the period from 1986 to 1998
(excluding firms operating in financial services). We find that, on average,
survey participants are more profitable, but growing at a slower rate, than
those in the sample of large Compustat firms. Specifically, the sample
average return on sales for survey participants is 17.8%, versus 15.7% for
the sample of large Compustat firms, and the average sales growth is 5.7%
versus 7.4%. This is consistent with our observation that the firms in our
sample are likely to be industry leaders (hence slightly more profitable)
and also large (hence with slightly slower growth). There is no reason why
this should dramatically skew the inferences from the sample.
14 Because the survey is based on benchmark jobs, it is possible that
nonstandard positions are excluded from the survey or added over time.
Companies may differ systematically as to the percentage of management
positions that are benchmark jobs, and this might bias our measure of the
span. Because, however, the positions reporting to the CEO are the most
763
is defined as the number of positions between the CEO and
the divisional CEO. In the survey, a division is defined as
“the lowest level of profit center responsibility for a business unit that engineers, manufactures and sells its own
products.”
We focus on the divisional CEO position (hereafter referred to as divisional manager) for two reasons: (i) it is the
position furthest down the hierarchy that is most consistently defined across firms, and (ii) it is informative about
the extent to which responsibility is delegated in the firm.
Figure A1 (at the end of the paper) displays an (edited)
example from the survey that demonstrates to participants
how to determine the number of reporting levels for each
position. The management reporting relationships are
clearly illustrated with the line of authority starting with the
CEO as the most senior position, moving down to the COO,
the group CEO, the divisional CEO, and finally the plant
manager as the most junior management position. In this
example, our measure of depth equals 2—there are two
positions between the CEO and the divisional manager.
Other positions that might be informative about the depth
of the hierarchy are group CEOs (managers with multiple
profit center responsibility) and plant managers (managers
with budget or cost center responsibility), but there are
limitations to using either. Group CEOs are defined on the
basis of their position in the hierarchy (proximity to CEO or
COO). Hence it is harder to infer facts about depth or
responsibility from their position. By contrast, divisional
managers are defined on the basis of their responsibility, and
hence we can infer more about hierarchies from where they
are placed. Unfortunately, though, the definition of plant
managers is not consistent across industries, especially
when one moves from manufacturing to service firms.15
The survey data are supplemented with information from
several other data sets (Compustat for financial and segment
information; Securities Data Company for mergers; Spectrum for institutional shareholdings; Gompers, Ishii, and
Metrick’s (2003) index, based on IRRC data, for governance information; and Directory of Corporate Affiliations
for year of founding). Whereas the survey is conducted in
April of each year and the organizational data describe the
firm in the year of survey completion, some statistics (including the number of employees in a division) represent
the end of the most recent fiscal year. To maintain consistency, we match the supplemental data sets using the year
prior to the year of the survey. Finally, not all variables are
senior positions and the primary focus of the survey over the period, we
expect the bias to be minimal.
15 The classic distinction between organizations that are organized by
function and by divisions with profit-center authority is less relevant in
this sample of firms. In fact, in this large sample of firms, pure functional
organizations appear to be uncommon. Using the reporting of divisional
and subsidiary data from the Directory of Corporate Affiliations, we were
able to categorize the structure of just over half of the sample firms. Only
one firm could be classified as a pure functional organization. In general,
information about organizational form did not add explanatory power to
our analysis.
764
THE REVIEW OF ECONOMICS AND STATISTICS
TABLE 2.—ORGANIZATIONAL SPAN (SPAN): NUMBER
OF
POSITIONS REPORTING
TO THE
Balanced Sample (N ⫽ 51)
Whole Sample (Unbalanced)
CHIEF EXECUTIVE OFFICER
Sample with 2 Consecutive Years
Mean
Median
STD
N
(firms)
Mean
Median
STD
1986
1987
1988
1989
1990
4.46
4.61
4.75
5.07
4.91
4
4
4
5
5
2.05
2.12
2.67
2.53
2.60
210
231
236
228
276
4.39
4.65
4.65
4.71
4.98
4
5
4
5
5
1.89
1.97
2.09
1.95
1.74
0.21
0.11
0.14
0.03
1.62
1.72
1.99
1.88
188
203
200
210
1991
1992
1993
1994
1995
4.81
4.89
5.01
5.38
5.65
4
5
5
5
5
2.96
2.50
2.24
2.45
2.54
289
290
304
298
288
5.25
4.96
5.53
5.82
6.47
5
5
5
5
6
2.08
2.12
2.10
2.15
2.64
⫺0.05
0.02
0.10
0.33
0.39
2.17
1.73
1.93
1.82
2.08
249
260
261
256
250
1996
1997
1998
5.46
6.10
6.79
5
6
6
2.56
2.94
3.90
280
248
213
6.31
7.08
8.16
6
6
7
2.32
2.75
4.02
⫺0.19
0.58
0.75
2.07
2.37
3.39
243
223
183
Average
N (firm-yr)
5.21
5
2.70
261
3,391
5.61
5
2.58
0.19
2.10
222
2,733
Year
Changes
Mean
Changes
STD
N
(firms)
Notes: Whole sample includes all firms in the sample. Balanced sample includes firms that appear in the sample over the 13-year period. Sample with 2 consecutive years includes all the firms in the sample for
the year and the year prior. Changes is the change in span between year t and year t ⫺ 1.
available for all positions, firms, and years, and due to
limitations in matching with the supplemental data sets, our
samples are smaller for some parts of the analysis.
III.
The Facts
A. Increasing Span
Having described the data and their sources, let us now
examine how firm hierarchies are changing over time. In
table 2, we describe how the number of positions reporting
directly to the CEO (that is, CEO Span) moves over the
period. The number of positions reporting has gone up from
a mean (median) of 4.46 (4) in 1986 to 6.79 (6) in 1998, an
increase of approximately 50%. One might worry that some
of the change is induced by changes in the composition of
our sample over time. If we restrict ourselves to the 51 firms
that appear throughout the 13 years of our panel, however,
the change is even more dramatic. From a mean (median) of
4.39 (4) it goes up to 8.16 (7), an increase of 86%. Alternatively, in the last three columns in table 2 we report the
average annual change in span for the firms that appear for
two consecutive years in the data set. Cumulating that
annual average change in span, we get a total of 2.42 over
the 13-year period.
Is this simply “hardwired”? Could increasing CEO span
reflect the natural growth of firms? No, because firms could
accommodate growth by adding layers to the hierarchy
rather than increasing the span of control, and because firms
have not grown significantly over this period. In fact, the
average size of the 51 firms appearing throughout, as measured by the number of employees, has fallen from 86,000
in 1986 to 74,000 in 1998 (see table 1A). In the unbalanced
panel, the size of firms is roughly constant over time—
approximately 48,000 in both 1986 and 1998. When we sort
firms into quartiles based on the growth in the number of
their employees over the sample, we do not find any clear
pattern in span across the quartiles (not reported in table).16
An obvious question is whether the growth in CEO reports
is a result of mergers—are firms simply stitched together at
the seams under a common CEO, and would the merger
wave account for our findings? To address this we drop from
the balanced sample all firms that undertook one or more
significant acquisitions (amounting to more than 20% of
assets in any year) in the previous 3 years. CEO reports still
increase from 4.4 in 1986 to 8.2 in 1998. We also drop from
the sample all firms that undertook significant acquisitions
at any time during the period covered. Again, CEO reports
increase from 4.4 in 1986 to 8.2 in 1998.
Another obvious question is whether the growth in CEO
reports is due to increases in diversification. In fact, the
average number of segments reported by Compustat (one
measure of diversification) for the balanced sample increases from 3.3 in 1986 to 3.9 in 1998 (table 1A). However, in a firm-fixed-effects regression of the number of
CEO reports on (the logarithm of) the number of employees,
the number of segments, and a trend variable, the coefficient
on the number of segments is insignificant, suggesting that
the increase in span is not primarily related to increases in
diversification.17
16 Firms may grow in size or complexity without adding more workers—
for example, through outsourcing. To evaluate this, we estimate all
regressions and replace firm and division employment with firm and
division sales, respectively. We find that our results are generally robust
using this alternative measure of size, with the rare exception noted.
17 One might even argue the reverse: the CEO plays a coordinating role,
so one would expect more reports to the CEO when there is more of a need
for coordination between various business segments, that is, when the
firm’s segments or divisions are more related. This conjecture too is not
borne out in the data. Using data on a division’s industry and the share of
employees in a two-digit industry within the firm, we calculate a Herfindahl index (HHI) for the firm’s presence in different industries as a more
refined measure of relatedness. In a firm-fixed-effects regression of the
THE FLATTENING FIRM
TABLE 3.—ORGANIZATIONAL SPAN: REPORTS
TO THE
CEO
BY
765
POSITION (BALANCED SAMPLE; N ⫽ 51)
Average Number
Corporate Staff Positions
Year
Intermediaries
Unit Heads
Chief
Chief
Chief
Chief
Group Manager
Division Manager
Information
Human
Financial General Strategic
Public
Operating Administrative
Officer
Resources Officer Counsel Planning Relations
Officer
Officer
Avg. No. Probability Avg. No. Probability
1986
1987
1988
1989
1990
0.020
0.078
0.039
0.020
0.039
0.373
0.451
0.490
0.490
0.510
0.667
0.686
0.686
0.706
0.667
0.667
0.667
0.686
0.725
0.725
0.275
0.255
0.255
0.255
0.294
0.196
0.235
0.294
0.333
0.431
0.549
0.529
0.549
0.510
0.588
0.392
0.353
0.392
0.314
0.333
1.026
0.897
0.789
0.947
0.970
0.434
0.432
0.417
0.407
0.419
0.205
0.340
0.213
0.205
0.229
0.052
0.097
0.063
0.073
0.084
1991
1992
1993
1994
1995
0.039
0.020
0.039
0.039
0.039
0.549
0.471
0.529
0.549
0.627
0.706
0.745
0.863
0.882
0.902
0.745
0.667
0.784
0.784
0.784
0.314
0.255
0.255
0.255
0.275
0.451
0.294
0.275
0.275
0.353
0.529
0.549
0.412
0.392
0.392
0.392
0.412
0.314
0.353
0.353
1.143
1.029
1.353
1.472
1.737
0.490
0.431
0.545
0.583
0.619
0.255
0.298
0.609
0.783
0.860
0.108
0.121
0.215
0.213
0.213
1996
1997
1998
0.039
0.078
0.176
0.667
0.706
0.647
0.961
0.941
0.902
0.843
0.902
0.961
0.235
0.235
0.392
0.314
0.412
0.569
0.412
0.431
0.451
0.275
0.275
0.294
1.721
2.051
1.733
0.556
0.670
0.606
0.581
0.535
0.953
0.179
0.159
0.314
Notes: Balanced sample includes firms that appear in the sample over the 13-year period. Positions are described in the appendix. For the group and divisional manager positions, the averages and probabilities
are calculated for the subset of firms reporting these positions. Probability is the fraction of group or divisional manager positions reported by the survey that report to the CEO.
As an aside, in what follows we have the option of
reporting data for the balanced panel of firms reporting
throughout or also reporting data for the unbalanced panel.
The balanced panel has the virtue of allowing comparisons
to be made for the same firms over time. It has the demerit
of focusing only on survivors and therefore introducing
potential biases. Fortunately, the patterns from the balanced
panel look qualitatively like those in the unbalanced panel.
Could the increased span be a result of the creation of
new positions such as chief information officer (CIO) or the
increasing importance of existing positions such as head of
human resources (HHR), who now join more traditional
positions such as chief financial officer in reporting directly
to the CEO? The data do not support this explanation.18 In
table 3, we report for the balanced panel the average number
of direct reports to the CEO from a particular position. Each
CEO had, on average, 0.02 CIOs and 0.37 HHRs reporting
in 1986. By 1998, each CEO had 0.18 CIOs and 0.65 HHRs
reporting. Thus these two positions account for only approximately 0.45 of the increased reports to the CEO. Where do
the rest of the reports (equal approximately to 8.16 ⫺
4.39 ⫺ 0.45 ⫽ 3.32) come from?
The answer seems to be that they come from traditionally
more junior positions. The average number of group managers reporting directly to the CEO went up from 1.03 in
1986 to 1.73 in 1998 (see table 3). The number of division
number of CEO reports on (the logarithm of) the number of employees,
HHI, and a trend variable, the coefficient on HHI is insignificant, suggesting that the increase in span cannot be explained by a greater need for
coordination.
18 The chief information officer (CIO, position 8 in the appendix) is
defined as the highest level of operating management over the combined
functions of programming, data processing, machine operation, and systems work related to data processing. Head of human resources (HHR,
position 7 in the appendix) is defined as the head of all human resources
with responsibility for establishing and implementing corporation-wide
policies.
managers reporting directly to the CEO went up from 0.21
in 1986 to 0.95 in 1998. Thus the increase in direct reports
from positions traditionally lower down in the organization
accounts for approximately 45% of what is unaccounted for
(0.70 ⫹ 0.74 ⫽ 1.44 of 3.32).19
The number of divisional manager positions reported by
survey participants has increased over time. So perhaps as
important as knowing the average number of group or
divisional managers who report to the CEO is knowing what
fraction of the group or divisional managers covered by the
survey report to the CEO. Call this the probability of
reporting to the CEO. For group managers this probability
increased slightly over the period, from 0.43 in 1986 to 0.61
in 1998 (see table 3). The probability that a divisional
manager reports to the CEO consistently trended upward
over the period, from 0.05 in 1986 to 0.31 in 1998.
Parenthetically, some traditionally senior positions have
also become closer to the CEO. Whereas 67% of CFOs
reported to the CEO in 1986, 90% did so in 1998. A similar
pattern is seen for the general counsel. Law and finance
seem to have become more important!
B. Decreasing Depth and Increasing Empowerment
Even though only some division managers report directly
to the CEO, the trend for them to be closer to the CEO is
more general. Table 4B column (ii) (balanced sample),
suggests that the average depth at which the division manager is located below the CEO (the number of managers
between the CEO and the division manager) has fallen, from
19 Some functions have increased considerably in importance. Only 0.2
public relations officers reported to the CEO in 1986, and the number
increased to 0.57 in 1998. Corporate research and development positions
and manufacturing positions account for approximately 0.20 of the remaining increase in the number of CEO reports.
766
THE REVIEW OF ECONOMICS AND STATISTICS
TABLE 4.—DESCRIPTIVE STATISTICS—FIRM
AND
BUSINESS-UNIT (DIVISION) CHARACTERISTICS (MEANS
Year
1986
1987
1988
1989
1990
(ii)
CHANGES)
B. Balanced Sample (N ⫽ 51)
Depth
(iii)
Division Size
(000s empz.)
(iv)
Division
Coverage
Sample with 2
Consecutive
Years:
Depth Changes
(Mean)
85.9
82.8
84.3
86.8
86.2
1.58
1.45
1.51
1.46
1.36
6.0
5.9
5.2
5.2
5.1
0.42
0.38
0.38
0.36
0.33
⫺0.06
0.00
⫺0.09
⫺0.05
289
290
304
298
288
86.9
83.2
81.6
81.8
81.5
1.33
1.35
1.20
1.19
1.25
4.2
4.4
4.6
5.1
4.8
0.35
0.33
0.33
0.37
0.33
⫺0.05
0.03
⫺0.04
⫺0.05
⫺0.01
0.43
0.38
0.39
280
248
213
79.6
75.4
73.8
1.30
1.41
1.18
5.5
2.8
4.7
0.37
0.34
0.40
0.01
⫺0.02
⫺0.06
0.43
261
82.7
1.36
4.9
0.36
⫺0.03
A. Whole Sample (Unbalanced)
(i)
Firm Size
(000s empz.)
AND
Depth
(iii)
Division Size
(000s empz.)
(iv)
Division
Coverage
(v)
N
(firms)
(i)
Firm Size
(000s empz.)
47.5
43.4
42.0
46.2
44.7
1.49
1.39
1.43
1.34
1.28
3.8
3.5
3.4
3.3
3.1
0.53
0.68
0.46
0.44
0.39
210
231
236
228
276
1991
1992
1993
1994
1995
42.1
41.3
38.9
41.1
39.3
1.26
1.29
1.19
1.08
1.09
3.1
3.1
2.8
3.1
3.4
0.40
0.37
0.37
0.42
0.41
1996
1997
1998
42.6
45.2
49.5
1.14
1.18
1.14
3.6
3.3
3.7
Average
43.1
1.26
3.3
(ii)
Notes: Whole sample includes all firms in the sample. Balanced sample includes firms that appear in the sample over the 13-year period. Sample with 2 consecutive years includes all the firms in the sample for
the year and the year prior. Firm (division) size is the number of employees in the firm (division) in thousands. Depth is defined as the number of positions between the CEO and the divisional manager (see figure
2 for an example). Division coverage is defined as the ratio of the number of employees under divisional manager positions sampled by the survey to the total number of employees in the firm. Depth changes is
defined as the depth in year t minus depth in year t ⫺ 1.
1.58 in 1986 to 1.18 in 1998, approximately 25%.20 Interestingly, the correlation between CEO Span and Depth is
significantly negative (correlation ⫽ ⫺0.27 for the whole
sample). Wider organizations are also less tall, so that, put in
a time series context, organizations are becoming flatter. In
what follows, we will focus on CEO Span as a measure of
organizational structure (because we believe it is more
comprehensively reported), though we will use Depth wherever appropriate.
Perhaps then the increasing number of reports to the CEO
reflects restructuring of business units: Perhaps profit center
responsibility has been taken away from smaller units, and
they are now part of a larger, more important unit whose
manager is, not surprisingly, closer to the CEO and now
may even report directly to him. Again, this hypothesis does
not seem consistent with the data. The average size of a
division (the lowest level of profit center responsibility) has
decreased from approximately 6,000 employees in 1986 to
4,700 employees in 1998 [see table 4B, column (iii)].
Of course, there may be a simpler explanation for our
findings. The survey is not exhaustive, except at the highest
levels in the organization. Perhaps, as the survey expanded
over time, it picked up lower, more obscure divisional
manager positions. This would explain why divisions are
getting smaller (but not why their depth is decreasing).
Nevertheless, even the premise is incorrect: the survey has
expanded in the number of divisional manager positions
reported but not in the fraction of the firm covered. For the
constant sample, we calculate the ratio of the total number
of employees under divisional manager positions sampled
20 In the last column in table 4 we report the average annual change in
depth for the set of firms that appear for two consecutive years in the data
set. Cumulating that average change in depth for this set of firms, we get
⫺0.03 over the 13-year period.
by the survey to the total number of employees in the firm.
As table 4B indicates, this ratio was 0.42 in 1986 and 0.4 in
1998. The coverage of the survey has not changed significantly.21
As yet, we cannot be sure whether the existing divisional
manager positions became closer to the CEO or whether
organizational change resulted in new divisional manager
positions that were closer to the CEO. For example, if large
firms started outsourcing more of their activities, new divisional managers might have been put in charge of units that
were not large as measured by personnel, but were only the
tip of a vast outsourced operation. It would not be surprising
then that these important managers would be closer to the
CEO.
One way to get at this is to follow the same divisional
manager position over time. From the annual surveys, we
identified which divisional manager positions were reported
multiple times over the years. Focusing only on these
positions, we regress attributes of the position (what its
depth is, whether it reports to the CEO) against the size of
the firm, the size of the division, year indicators, and an
indicator for the position. These regressions should be
viewed as attempts to establish partial correlations rather
than as implied causal relationships. Significant coefficient
estimates on the year indicators would only suggest that,
keeping the other attributes of a position approximately
constant, its place in the organizational hierarchy did change
over time.
The regression estimates are reported in table 5. The
standard errors for the reported estimates are clustered at the
firm level, addressing the concern that division observations
21 A similar conclusion is reached if one examines the coverage of group
positions reported (results available on request from the authors).
THE FLATTENING FIRM
TABLE 5.—MEASURES
Independent variable
log(Division Employees)
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
Constant
Observations
Number of divisions
R-squared
OF
767
EMPOWERMENT—DIVISION-FIXED-EFFECTS REGRESSIONS
DDEPTH
(i)
CEORPT
(ii)
OFFICER
(iii)
⫺0.074***
(0.020)
⫺0.052
(0.040)
⫺0.034
(0.058)
⫺0.082
(0.061)
⫺0.152**
(0.060)
⫺0.167***
(0.060)
⫺0.132*
(0.070)
⫺0.195***
(0.069)
⫺0.254***
(0.074)
⫺0.236***
(0.076)
⫺0.270***
(0.077)
⫺0.285***
(0.081)
⫺0.302***
(0.088)
2.077***
(0.143)
0.019**
(0.008)
0.014
(0.015)
0.012
(0.020)
0.020
(0.020)
0.040*
(0.024)
0.049**
(0.025)
0.047*
(0.024)
0.067**
(0.027)
0.095***
(0.029)
0.095***
(0.033)
0.088***
(0.032)
0.098***
(0.038)
0.071*
(0.038)
⫺0.066
(0.057)
0.034***
(0.011)
0.015
(0.012)
0.018
(0.016)
0.006
(0.017)
0.017
(0.020)
0.030
(0.020)
0.027
(0.020)
0.033
(0.024)
0.041
(0.026)
0.060*
(0.034)
0.053
(0.035)
0.052
(0.041)
0.045
(0.036)
⫺0.036
(0.077)
10393
2360
0.73
10428
2370
0.58
10428
2370
0.77
Dependent variables are DDEPTH (number of positions between CEO and divisional manager), CEORPT (divisional manager position reports to CEO), and OFFICER (incumbent in divisional manager position
is corporate officer).
Notes: Includes all divisions in the sample that appear for at least two years. All specifications include division fixed effects. All variables have been Winsorized at the 99th percentile. DDEPTH is defined as
the number of positions between the CEO and the specific divisional manager position. CEORPT is a dummy variable equal to 1 if the divisional manager position reports directly to the CEO and 0 otherwise.
OFFICER is a dummy variable equal to 1 if the incumbent in the divisional manager position is a corporate officer and 0 otherwise. log(Division Employees) is defined as the log of the number of employees in
the division. All specifications report robust standard errors by clustering at the firm level. ***/**/* represent significance at the 1%/5%/10% level.
may not be independent across divisions within a firm.22 In
column (i), the dependent variable is the depth of the
specific position (DDEPTH). We find negative coefficients
on all year indicators, with a trend of increasingly negative
coefficients over time. That is, division depth is decreasing,
and divisional manager positions are getting closer to the
top. In column (ii), the dependent variable is 1 if the
position reports to the CEO directly and 0 otherwise. We
find that the probability of reporting to the CEO increases
over time, as the year indicators become larger over the
period. Also, the number of employees under a particular
divisional manager position trends downward slowly (approximately 1% every year). This suggests that even though
the structure of the division has not changed drastically over
22 Clustering standard errors at the firm level instead of the division level
addresses the possibility that firms have certain rules (or standards) by
which they place all of their managers. In this case, the different positions
in a firm are not independent. Clustering by firm is a more conservative
test and particularly important if there is a lot of covariance between the
positions of managers in the same firm. In addition to clustering by firm,
we estimate regressions that adjust standard errors for serial correlation
(AR1) and heteroskedasticity across division manager positions. Because
the statistical significances of the coefficients are similar for the two
approaches, we choose to report standard errors clustered at the firm level.
time, its head has moved nearer the top. The organizational
hierarchy is indeed becoming flatter.
Finally, a direct measure of responsibility is whether the
holder of a position is designated an officer of the corporation: officers of the corporation are determined by both the
individual’s authority and the nature and extent of the
individual’s duties.23 In column (iii) of table 5, the dependent variable is whether the divisional manager is designated an officer. The year coefficients exhibit a broadly
increasing trend over the entire sample, with the year
coefficients averaging 0.014 in the first third of the sample
years (1987–1890), and 0.053 in the last third (1995–1998).
Although only one of the year coefficients in the last third is
significantly different from 0, collectively they are greater
23 The term “officer” is defined by the Securities and Exchange Commission in Section 240.16 (rules governing insider trading) as “an issuer’s
president, principal financial officer, principal accounting officer (or, if
there is no such accounting officer, the controller), any vice-president of
the issuer in charge of a principal business unit, division or function (such
as sales, administration, or finance), any other officer who performs a
policy-making function.”
768
THE REVIEW OF ECONOMICS AND STATISTICS
than 0 at the 11% level.24 By contrast, the year coefficients
in the initial third of the sample are not statistically different
from 0. The incumbent in the divisional manager position
has become somewhat more likely to be designated an
officer over time. Authority and responsibility are indeed
moving further down.
C. Delayering
That the CEO is getting more directly connected—increasing span, reduced distance from managers—is consistent with anecdotal evidence that organizations have been
getting rid of entire layers of middle management. In general, it is hard to find direct evidence of this at the level of
detail our data set offers on reporting relationships—the
mere fact that positions disappear does not mean that
reporting has become more direct, for other positions could
place themselves in the middle.25
Not only do our data suggest that reporting has become
more direct (for instance, that more division managers now
report directly to the CEO), but they also suggest that the
CEO is becoming more directly connected precisely because of the elimination of intermediate positions: Consider
the position of the COO, who has historically served as an
intermediary between the CEO and the rest of the organization. As table 3 indicates, the average number of COO
reports to the CEO per firm has fallen from 0.55 to 0.45 over
the same period. The position of chief administrative officer
(CAO) also seems to exhibit a similar decline. The decline
in COO and CAO positions that report to the CEO is
primarily because these intermediate positions are being
eliminated, and not necessarily because these officers have
less access to the CEO. Conditional on a firm having a
COO, the percentage of COOs that reported to the CEO
didn’t change over the period (very close to 100%). This
suggests that the decline in COO reports to the CEO is due
to the position being eliminated in the sample firms.
In Table 6, column (i), we return to the unbalanced sample
and regress CEO Span against a constant, firm size (the log of
the number of employees in the firm), firm, and year indicators.
The trend in the coefficient estimates on the year indicators is
significantly positive. CEO Span increases, on average, by
approximately 0.16 every year. In column (ii), we also include
an indicator for whether the firm has a COO and another
indicator for whether it has a CAO. The coefficients on the year
indicators fall slightly. Interestingly, the coefficient on the
presence of a COO is negative, statistically significant, and
24 This is one situation where including division sales, as we have done,
instead of division employees reduces the significance of the coefficients.
Also, clustering by firm greatly reduces significance relative to clustering
by division. When we cluster by division, seven of the year indicators are
statistically significant.
25 Earlier work has inferred reporting relationships from organizational
positions (managers in lower layers are assumed to report to managers in
the immediate higher layer). In this case, the elimination of some, but not
all, positions in intermediate layers would not allow us to conclude that
there is a change in reporting relationships.
large (⫺0.96). Assuming the COO always reports to the CEO,
this coefficient suggests his presence reduces the number of
CEO reports, because an average of 1.96 managers who would
otherwise report to the CEO now report to him. In other words,
the COO is truly an intermediary.26
In column (iii) we regress Depth on firm size and on firm
and year indicators, and we find that the coefficients on the
year indicators become increasingly negative over the period. Column (iv) suggests the presence of intermediaries
like the COO and the CAO, increase the average depth at
which division managers are positioned. If the COO stood
between the CEO and all managers, the coefficient on the
COO indicator would be 1. That it is lower suggests some
divisional managers do not report via the COO. Parenthetically, note that the coefficient on firm size is positive,
suggesting that growing firms seem to have greater depth.
Although the coefficients on the year indicators fall when
we include indicators for the presence of these positions,
they do not become insignificant. Thus the elimination of
the COO and CAO positions accounts for part, but not all,
of the trend. The flattening of organizations is more than the
elimination of just a few key intermediate positions.
D. The Correlation with Wages
Are increasing span and decreasing depth simply changes
on paper with no real consequences whatsoever? Does the
ostensible proximity to the CEO simply reflect a greater
desire on the part of managers for status, with no greater
increase in real access? Evidence that more division managers are becoming officers suggests that organizations are
changing in meaningful ways. And one strong piece of
evidence suggests that these changes are not all form without any function: they seem to be accompanied by systematic changes in pay.27
The data set we have has extensive data on compensation.
We would like to see if the flattening of the hierarchy we
have described has any correlation with pay patterns. To
understand this, we examine the pay of divisional managers,
who could be positioned anywhere from just below the CEO
to far away.
In Table 7, we report how various attributes of the pay
structure for firms vary as depth decreases. The first aspect
of pay we consider is the divisional manager’s salary and
bonus. We regress the logarithm of this measure against the
number of employees in the firm, the number of employees
in the division, the depth at which the division manager is
placed, and year indicators. The OLS estimate for the
26 By contrast, the presence of a CAO increases CEO reports, but by less
than 1. Because the CAO also typically reports directly to the CEO, the
coefficient estimate of 0.344 suggests that the CAO also intermediates
between lower positions and the CEO, but typically has fewer direct
reports than the COO.
27 Larcker, Lambert, and Weigelt (1993) evaluate pay differentials between different positions in a firm’s hierarchy. However, they have no
measure of CEO span or the number of levels between the CEO and
division managers.
THE FLATTENING FIRM
TABLE 6.—ORGANIZATIONAL SPAN
769
DEPTH—FIRM FIXED-EFFECTS REGRESSIONS
AND
SPAN
Independent Variable
log(Employees)
(i)
(ii)
(iii)
(iv)
⫺0.172
(0.234)
0.307***
(0.081)
0.291**
(0.138)
0.320*
(0.170)
0.569***
(0.172)
0.544***
(0.178)
0.462**
(0.188)
0.562***
(0.194)
0.661***
(0.190)
1.077***
(0.207)
1.407***
(0.213)
1.303***
(0.223)
1.776***
(0.244)
2.349***
(0.272)
4.816***
(0.695)
⫺0.163
(0.237)
⫺0.964***
(0.148)
0.344**
(0.169)
0.264**
(0.133)
0.216
(0.164)
0.462***
(0.164)
0.432**
(0.170)
0.326*
(0.178)
0.463**
(0.185)
0.530***
(0.182)
0.917***
(0.200)
1.242***
(0.207)
1.154***
(0.220)
1.644***
(0.240)
2.183***
(0.269)
5.244***
(0.702)
⫺0.092**
(0.043)
⫺0.053
(0.053)
⫺0.143**
(0.058)
⫺0.183***
(0.063)
⫺0.215***
(0.062)
⫺0.139**
(0.065)
⫺0.167**
(0.066)
⫺0.219***
(0.067)
⫺0.231***
(0.071)
⫺0.283***
(0.069)
⫺0.270***
(0.070)
⫺0.345***
(0.079)
0.498*
(0.253)
0.294***
(0.072)
0.457***
(0.039)
0.038
(0.043)
⫺0.085**
(0.039)
⫺0.011
(0.048)
⫺0.102*
(0.054)
⫺0.141**
(0.058)
⫺0.152***
(0.057)
⫺0.101*
(0.057)
⫺0.098
(0.060)
⫺0.142**
(0.061)
⫺0.146**
(0.066)
⫺0.208***
(0.065)
⫺0.193***
(0.065)
⫺0.251***
(0.073)
0.266
(0.222)
3264
369
0.49
3264
369
0.52
2381
323
0.65
2381
323
0.71
COO
CAO
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
Constant
Observations
Number of firms
R-squared
DEPTH
Dependent variables are SPAN (number of positions reporting to CEO) and DEPTH (firm average number of positions between the CEO and the divisional manager).
Notes: Includes all firms in the sample that appear for at least two years. All specifications include firm fixed effects. All variables have been Winsorized at the 99th percentile. log(Employees) is defined as the
log of the number of employees in the firm. COO and CAO are dummy variables equal to 1 if the firm reports a chief operating officer and chief administrative officer, respectively. All specifications report robust
standard errors by clustering at the firm level. ***/**/* represent significance at the 1%/5%/10% level.
coefficient of depth [table 7, column (i)] is negative and
significant, suggesting that for each additional layer between the CEO and the divisional manager, the log of the
latter’s salary and bonus falls by 0.14, which is 29.4% of its
standard deviation. The coefficient continues to be negative
when we include fixed effects for the position [table 7,
column (ii)], suggesting that specific divisional manager
positions that are moving closer to the CEO over time get
paid more.
We find similar results when the dependent variable is the
ratio of the division manager’s long-term incentive pay to
the value of salary and bonus (typically stock and stock
options).28 The OLS estimate [table 7, column (iii)] suggests
28 The value of long-term incentive pay is computed by Hewitt. Stock
options are valued using a modified version of the Black-Scholes model
that takes into account vesting and termination provisions in addition to
the standard variables of interest rates, stock price volatility, and dividends. As is standard practice among compensation consulting firms, the
other components of long-term incentives are valued using an economic
valuation similar to the Black-Scholes that takes into account vesting,
termination provisions, and the probability of achieving performance
goals.
that long-term incentive pay for divisional managers goes
down from a mean of 43.4% of salary and bonus to 37.1%
for an additional layer between the CEO and the divisional
manager. Including division fixed effects does not change
the estimate significantly.29
All this suggests that pay and incentives are adjusting to
the change in organizational structure. Even controlling for
the size of the division and the firm, division managers are
paid more as they move closer to the CEO. This is therefore
not the traditional Calvo-Wellisz (1979) or Rosen (1982)
effect—it is not that these managers are becoming more
important at the margin because they control larger operations.
Instead, it may well be that they are becoming more
important because their decision-making is subject to less
close oversight by intermediaries (though to more direct
oversight by the CEO). In fact, as Aghion and Tirole (1997)
argue, greater span may be a way for the CEO to commit to
29 See Wulf (2006) for additional analysis of the relationship between
divisional manager incentives and organizational structure.
770
THE REVIEW OF ECONOMICS AND STATISTICS
TABLE 7.—DIVISIONAL MANAGER PAY
AND
DEPTH: OLS
AND
DIVISION FIXED-EFFECTS REGRESSIONS
Divisional Manager
Salary ⫹ Bonus
Independent
Variable
log(Employees)
log(Division Employees)
DDEPTH
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
Constant
Observations
R-squared
Divisional Manager
LT Incentives
OLS
(i)
Division
Fixed Effects
(ii)
OLS
(iii)
Division
Fixed Effects
(iv)
0.059***
(0.023)
0.128***
(0.013)
⫺0.141***
(0.026)
0.080***
(0.016)
0.162***
(0.023)
0.182***
(0.024)
0.205***
(0.027)
0.213***
(0.026)
0.279***
(0.027)
0.291***
(0.029)
0.414***
(0.035)
0.431***
(0.039)
0.429***
(0.043)
0.526***
(0.048)
0.531***
(0.039)
11.236***
(0.087)
0.060*
(0.032)
0.083***
(0.011)
⫺0.065***
(0.012)
0.078***
(0.015)
0.139***
(0.020)
0.175***
(0.018)
0.202***
(0.022)
0.216***
(0.021)
0.281***
(0.022)
0.314***
(0.025)
0.402***
(0.026)
0.458***
(0.029)
0.464***
(0.033)
0.545***
(0.035)
0.539***
(0.037)
11.427***
(0.124)
0.046***
(0.015)
0.042***
(0.008)
⫺0.059***
(0.015)
0.035*
(0.019)
0.046
(0.029)
0.058**
(0.028)
0.118***
(0.030)
0.098***
(0.027)
0.104***
(0.030)
0.092***
(0.026)
0.166***
(0.033)
0.222***
(0.037)
0.266***
(0.038)
0.382***
(0.045)
0.362***
(0.045)
⫺0.061
(0.061)
0.084*
(0.047)
0.038***
(0.008)
⫺0.053***
(0.012)
0.022
(0.023)
0.016
(0.028)
0.052*
(0.030)
0.115***
(0.030)
0.105***
(0.029)
0.108***
(0.031)
0.102***
(0.031)
0.160***
(0.039)
0.223***
(0.042)
0.304***
(0.044)
0.373***
(0.050)
0.351***
(0.056)
⫺0.164
(0.150)
9915
0.43
9915
0.86
9915
0.17
9915
0.66
Dependent variables are (log) divisional manager salary plus bonus and divisional manager LT incentives (ratio of value of long-term incentive pay to salary plus bonus).
Notes: Includes all divisions in the sample that appear for at least two years. All variables have been winsorized at the 99th percentile. Divisional manager LT incentives is defined as the ratio of the value of
long-term incentive pay for divisional managers to the sum of the salary and bonus. Long-term incentive pay includes restricted stock, stock options, and other forms of long-term incentives (such as performance
units, performance share plans, and phantom stock). Refer to footnote 27 in the text for the consulting firm’s valuation of long-term incentives. DDEPTH is defined as the number of positions between the CEO
and the specific divisional manager position. log(Employees) is the log of the number of employees in the firm. log(Division Employees) is the log of the number of employees in the division. All specifications report
robust standard errors by clustering at the firm level. ***/**/* represent significance at the 1%/5%/10% level.
light monitoring and increased delegation of authority, as he
does not have enough time for detailed scrutiny of all his
subordinates. That authority is being delegated could also
explain why long-term incentive pay is increasing for those
who are moving closer to the top.
The reader should not, however, conclude that flattening
is simply decentralization of CEO authority. A broader span
of control also suggests the CEO’s direct involvement in
decisions across a greater number of organizational units.
Let us now turn to possible explanations.
IV.
Possible Explanations for the Flattening of Firms
A. Increased Competition in Product Markets
One set of possible explanations has to do with the more
competitive environment in product markets. Deregulation
and increased trade has enhanced product market competition over the last few decades. Not only has the required
speed of response for firms increased, it has put a premium
on employee competence and creativity. The tall hierarchies
of the past may no longer be as effective.
One reason may simply be because decisions need to be
taken more quickly to take advantage of fleeting opportunities in the marketplace—this is suggested by the GE CEO
Jeffrey Immelt’s desire, cited in the introduction, for “faster
decision making and execution.” It takes time for each
managerial layer to give approval to a decision. As the speed
increases with which a final decision is needed to avail of
opportunities, either a number of opportunities are forgone
with the attendant loss of value, or final decisions are
delegated further down a hierarchy with attendant loss of
top management control. Ceteris paribus, Williamson
(1967) or Calvo and Wellisz (1978) would suggest that
organizations would tend to become flatter in this environment.
THE FLATTENING FIRM
Also, greater competition may increase the complexity of
the decisions that have to be made as well as the variety of
data that impinge on the decision. Tall hierarchies with
intermediate managers micromanaging the work of operational managers may stifle initiative (see, for example,
Aghion & Tirole, 1997). Also, information may be hard to
convey up a hierarchy with the necessary detail and color,
thus reducing managers’ incentive to collect it (Stein, 2002).
Thus tall hierarchies may become dysfunctional, with top
managers not having the information to make the right
decisions and operational managers not having the incentive
to make them.
Finally, as the development of financial markets has increased access to physical capital, and as human capital becomes more important to a firm’s comparative advantage (see,
for example, Dessein, 2002; Rajan & Zingales, 2000; Roberts
& Van den Steen, 2000), tall hierarchies may lead to top
management losing the residual rights of control. In Grossman
and Hart (1986), subordinate managers are said to be controlled by virtue of top management’s ownership (actual or
delegated) of physical assets. If physical assets become relatively unimportant, ownership becomes less effective as a
means of organizational control. Tall hierarchies become less
viable. Instead, as Rajan and Zingales (2001) argue in their
development of the Grossman-Hart framework, top management has to build up control. It does this by establishing direct
contact with lower-level managers (that is, flattening the firm)
and getting them to make human-capital-specific investments
vis-à-vis top management.30 Thus the human-capital-intensive
firm is held together by a web of human-capital-specific investments, which are made possible by the flatter hierarchy. In
addition, employees get the promise of substantial ownership
rights, especially at the top, giving them an incentive to stay
with the firm despite having many competitors for the top
positions.
Tests of the broad hypothesis could be developed. For
instance, using measures of the timing and extent of deregulation of different industries, we can check whether deregulation led to flattening. Of course, it would be important to check
that the deregulation was not anticipated. Similarly, one can
look at import penetration as a measure of enhanced competition. A preliminary analysis suggests that the trend of decrease in depth is more pronounced in industries where the
percentage change in imports was higher between the years
1990 and 1994, based on data documented in Feenstra (1996).
However, more careful analysis that corrects for the endogeneity of imports (imports are more likely in industries that are
not efficient), and that distinguishes between imports of final
30 The entire organization becomes flatter in Rajan and Zingales
(2001)—senior managers cannot risk giving lower managers too many
subordinates, lest they become too independent. Hierarchies become
wider, middle managers are eliminated, and the firm bifurcates into top
management, who are owners/partners and can be trusted with command
over many subordinates, and worker/managers, who cannot be trusted till
they have served time in the firm [see Rebitzer and Taylor (1997) for an
early study of the structure of law firms suggesting this pattern].
771
goods and imports of inputs, will be required for results to be
convincing. We leave such analysis to future work.
B. Improvements in Corporate Governance
An important class of explanations has to do with agency
costs. In particular, a number of theorists have hypothesized
that management, left to itself, might want to expand its turf
and sense of worth (see, for example, Jensen, 1986; Jackall,
1988; Osterman, 1996; Parkinson, 1958; Useem, 1996) by
hiring legions of useless middle managers. This may have
been possible in the past when management was inadequately monitored. Equivalently, if the external governance
of the firm is poor, firms may not fire incompetent managers, but simply hire new ones to do their job.
If firms developed tall, overstaffed hierarchies because of
empire building, then improvements in corporate governance
could explain the trend toward flatter organizations. Governance benefited in the 1980s from the wave of hostile takeovers, which stepped up pressure on the large firms that
constitute our sample. The corporate raider, Carl Icahn, described his goal as eliminating “layers of bureaucrats reporting
to bureaucrats.”31 In the 1990s, large institutional investors
replaced the hostile takeover as the source of governance (see,
for example, Kaplan, 1996). Useem (1996) suggests that the
growing dominance of institutional investors in the stock
market has forced structural change in corporations: the elimination of layers of middle management and the restructuring
of firms into more autonomous business units.
However, when we regress the depth of a firm’s organizational structure on crude proxies for the extent of governance
pressure on the firm, such as the extent of institutional shareholding in that firm (lagged 1 year) or a measure of the strength
of outside governance compiled by Gompers, Ishii, and Metrick (2003), we find little systematic relationship (estimates
available from authors). Moreover, if greater depth were a
symptom of empire building, it should hurt the firm’s value,
and we should see a negative correlation between Depth and
the market-to-book ratio. In a regression of a firm’s market-tobook ratio on firm size, number of segments (diversified firms
typically have a lower market-to-book ratio), Depth, year
dummies, and fixed effects for the firm, we find no relationship
between the market-to-book ratio and Depth.
Of course, one explanation of these findings is that the
market for corporate control worked perfectly. Some firms
made timely changes in their organizational structure, while
those that did not swiftly attracted external governance
pressure, which forced them to change. As a result, we
might find no relationship between organizational structure
and measures of governance. Similarly, because all firms
adjusted in a timely fashion to the optimal extent, whether
voluntarily or not, there might be no relationship between
depth and firm value. From all this, we can only conclude
31
Quoted in Osterman (1996, p. 17).
772
THE REVIEW OF ECONOMICS AND STATISTICS
that more work is needed to establish that better corporate
governance has led to flatter hierarchies.
C. Information Technology
Another quite plausible explanation for the flattening of
hierarchies is changes in information technology. In a classic article, Leavitt and Whisler (1958) predicted that the
introduction of information technology into organizations
would reduce the number of middle managers because their
information gathering and coordinating role would be eliminated. Though there is some evidence that the introduction
of information technology leads to smaller firms (see, for
example, Brynjolfsson et al., 1994), others have argued that
the introduction of information technology increases the
richness of data to be analyzed and acted upon, and therefore creates more of a role for middle managers [for an
excellent discussion, see Pinsonneault and Kraemer (1997)].
As recent models suggest, theoretical predictions of the
effect of improvements in information technology on organizational change depend on whether information technology
reduces the cost of communication or whether it increases the
capability of lower managers to access information to make
decisions (see Garicano, 2000). According to his theory, increases in the use of information technology increase the span
of control for managers, but the effect on the depth of hierarchies is more ambiguous (predictions depend on whether the
technology primarily eases communication or access to information). Thus a careful test of information-based theories
requires much more detailed knowledge of the kind of work
done in a position. When combined with the difficulty of
obtaining good proxies on the extent of use of information
technology, this leads us to think tests are again best left for
future work.32
V.
Conclusion
In conclusion, we have unearthed a new set of facts about
the changing nature of corporate hierarchies: firms are
becoming flatter, the CEO span is broader, intermediate
managers are being dispensed with, and divisional managers are getting more authority, higher pay, and greater
incentive pay as they come closer to the CEO. Although our
findings suggest that corporate hierarchies are becoming
flatter, this phenomenon cannot be characterized simply as
centralization or decentralization. On the one hand, the CEO
is getting directly connected deeper down in the organization across a greater number of organizational units, a form
of centralization. On the other hand, decision-making authority and incentives are also being pushed further down, a
form of decentralization. We have offered a set of explanations for these facts. Testing these and other explanations
offers ample scope for future work.
32 For an illustration of the difficulty in disentangling the complex
relationship between IT and work practice, see Bresnahan, Brynjolfson,
and Hitt (2002).
In sum, then, this paper unearths interesting patterns of
change in firm hierarchies and provides some evidence that
this has to do with changes in the nature of the activities
being governed. This supports a central theme in the literature stemming from Coase (1937) and Williamson (1975,
1985) and suggests that developments in that literature can
offer valuable tools to understand organizational structures.
REFERENCES
Aghion, Philippe, and Jean Tirole, “Formal and Real Authority in Organizations,” Journal of Political Economy, 105 (1997), 1–27.
Baker, G., M. Gibbs, and B. Holmstrom, “The Internal Economics of the
Firm: Evidence from Personnel Data,” The Quarterly Journal of
Economics 109:4 (1994), 881–919.
Bresnahan, Timothy F., Erik Brynjolfsson, and Lorin M. Hitt, “Information Technology, Workplace Organization, and the Demand for
Skilled Labor: Firm-Level Evidence,” The Quarterly Journal of
Economics 117:1 (2002), 339–376.
Brynjolfsson, Erik, Thomas W. Malone, Vijay Gurbaxani, and Ajit Kambil, “Does Information Technology Lead to Smaller Firms?” Management Science 40:12 (1994), 1628–1645.
Calvo, G., and S. Wellisz, “Supervision, Loss of Control, and the Optimal
Size of the Firm,” Journal of Political Economy 87 (1978), 943–
952.
“Hierarchy, Ability, and Income Distribution,” Journal of Political Economy 87:5 (1979), 991–1010.
Child, J., “Predicting and Understanding Organization Structure,” Administrative Science Quarterly 18 (1973), 168–185.
Coase, Ronald, “The Nature of the Firm,” Economica 4:4 (1937), 386–405.
Dessein, Wouter, “Authority and Communication in Organizations,” Review of Economic Studies 69:4 (2002), 811–838.
Feenstra, Robert C., “U.S. Imports, 1972–1994: Data and Concordances,”
NBER working paper no. WSS15 (1996).
Garicano, Luis, “Hierarchies and the Organization of Knowledge in
Production,” Journal of Political Economy 108:5 (2000), 874–904.
Gompers, Paul A., Joy L. Ishii, and Andrew Metrick, “Corporate Governance and Equity Prices,” The Quarterly Journal of Economics
118:1 (2003), 107–155.
Grossman, Sanford J., and Oliver D. Hart, “The Costs and Benefits of
Ownership: A Theory of Vertical and Lateral Integration,” Journal
of Political Economy 94:4 (1986), 691–719.
Holmstrom, Bengt, and Steven N. Kaplan, “Corporate Governance and
Merger Activity in the United States: Making Sense of the 1980s and
1990s,” The Journal of Economic Perspectives 15:2 (2001), 121–145.
Jackall, Robert, Moral Mazes: The World of Corporate Managers (New
York: Oxford University Press, 1988).
Jensen, Michael C., “Agency Costs of Free Cash Flow, Corporate Finance,
and Takeovers,” American Economic Review 76 (1986), 323–329.
Kaplan, S., “Are We All Becoming Henry Kravis Now? The Evolution of US
Corporate Governance,” University of Chicago working paper (1996).
Lambert, R., Larcker, D., and Weigelt, K., 1993, “The Structure of
Organizational Incentives,” Administrative Science Quarterly 38:
438–461.
Leavitt, H.H., and T.I. Whisler, “Management in the 1980s,” Harvard
Business Review (November–December 1958), 41–48.
Osterman, P. Broken Ladders: Managerial Careers in the New Economy
(New York: Oxford University Press, 1996).
Parkinson, C. Northcote, Parkinson’s Law or the Pursuit of Progress
(London: J. Murray, 1958).
Pinsonneault, Alain, and Kenneth L. Kraemer, “Middle Management
Downsizing: An Empirical Investigation of the Impact of Information Technology,” Management Science 43:5 (1997), 659–679.
Powell, Walter, “Neither Market nor Hierarchy: Network Forms of Organization” (pp. 295–336), in Barry Straw and Lawrence Cummings
(Eds.), Research in Organizational Behavior (Greenwich, CT: JAI
Press, 1990).
“The Capitalist Firm in the Twenty-First Century: Emerging
Patterns in Western Enterprise” (pp. 33–68), in Paul Di Maggio
(Ed.), The Twenty-First Century Firm (Princeton, NJ: Princeton
University Press, 2001).
THE FLATTENING FIRM
Pryor, Frederic L., “The Economics of the Good Society: The Variety of
Economic Arrangements,” Journal of Economic Literature 38:3
(2000), 681–683.
Pugh, D.S., D.J. Hickson, C.R. Macdonald, K.M. Turner, C. Turner, and
T. Lupton, “Dimensions of Organization Structure,” Administrative
Science Quarterly 13 (1968), 65–105.
Rajan, R.G., and L Zingales, “The Governance of the New Enterprise”
(pp. 201–227), in Xavier Vives (Ed.), Corporate Governance,
Theoretical and Empirical Perspectives (Cambridge University
Press, 2000).
“The Firm as a Dedicated Hierarchy: A Theory of the Origins and
Growth of Firms,” The Quarterly Journal of Economics 116:3
(2001), 805–851.
Rebitzer, John B., and L.J. Taylor, “When Knowledge Is an Asset:
Explaining the Organizational Structure of Large Law Firms,” MIT
working paper (1997).
Roberts, John, and Eric Van den Steen, “Shareholder Interests, Human
Capital Investment and Corporate Governance,” Stanford GSB
working paper (2000).
Rosen, Sherwin, “Authority, Control, and the Distribution of Earnings,”
The Bell Journal of Economics 13:2 (1982), 311–323.
Scott, Elizabeth D., K.C. O’Shaughnessy, and Peter Cappelli, “Management Jobs in the Insurance Industry: Organizational Deskilling and
Rising Pay Inequity” (pp. 126–154), in Paul Osterman (Ed.),
Broken Ladders: Managerial Careers in the New Economy (New
York and Oxford: Oxford University Press, 1996).
Stein, Jeremy C., “Information Production and Capital Allocation: Decentralized versus Hierarchical Firms,” Journal of Finance 57:5
(2002), 1891–1921.
Useem, M. “Corporate Restructuring and the Restructured World of
Senior Management” (pp. 23–54), in Broken Ladders: Managerial
Careers in the New Economy (New York and Oxford: Oxford
University Press, 1996).
Williamson, Oliver E., “Hierarchical Control and Optimum Firm Size,”
Journal of Political Economy 75:2 (1967).
Markets and Hierarchies: Analysis and Antitrust Implications
(New York: Free Press, 1975).
“The Modern Corporation: Origins, Evolution, Attributes,” Journal of Economic Literature 19:4 (1981).
The Economic Institutions of Capitalism (New York: Free Press,
1985).
Wulf, Julie, “Authority, Risk, and Performance Incentives: Evidence from
Division Manager Positions inside Firms,” Journal of Industrial
Economics, forthcoming (2006).
4.
5.
6.
7.
8.
9.
10.
11.
773
data processing, public relations, and long-range planning and
business development.
Chief financial officer (CFO). Functional head responsible for all
financial operations of the corporation. Has responsibility for both
the treasury and accounting functions. Indicate whether responsibilities also include data processing, investor relations, internal
audit, and taxes.
Long-range planning & business development. Functional head
responsible for developing and obtaining agreement on overall
corporate strategy to enhance sales and profits. Recommends the
allocation of resources to existing businesses, acquisitions of new
businesses, and disposition of existing businesses.
General counsel. The head of all legal affairs of the company.
Responsible for, or may be, corporate secretary; supervises outside
legal counsel.
Human resources. Head of all human resources with responsibility
for establishing and implementing corporation-wide policies.
Chief information officer (CIO). The highest level of operating
management over the combined functions of programming, data
processing, machine operation, and systems work related to data
processing.
Public relations. Functional head responsible for the development
and dissemination of favorable persuasive material in order to
promote goodwill, develop credibility, and create a favorable
public image for the company.
Group chief executive (or group manager). The highest authority
in the group. A group is the highest level of multiple profit center
linking the corporate CEO or COO directly to two or more single
profit center units (divisions).
Division chief executive (or divisional manager). The highest
authority in the division. A division is the lowest level of profit
center responsibility for a business unit that engineers, manufactures, and sells its own products.
FIGURE A1.—EXAMPLE OF REPORTING LEVELS, DEPTH, SPAN,
DESCRIPTIONS OF TYPES OF ORGANIZATIONAL UNITS
AND
APPENDIX
Position Descriptions
1. Chief executive officer (CEO). The highest executive authority in
the corporation. Reports to the board of directors. May also be
chairman or president.
2. Chief operating officer (COO). The corporation’s second in command, provided the person’s span of responsibility is as broad or
almost as broad as the chief executive’s, and provided he or she
has line rather than staff or advisory responsibility. This person
may be the president if the chief executive officer is the chairman
of the board.
3. Chief administrative officer (CAO). Functional head responsible
for the administration of two or more major, nonrelated corporate
staff functions such as finance, human resources, law, purchasing,
Types of organizational units:
●
●
●
●
A corporate unit is the highest management organization level of the parent company, responsible
for its overall direction.
A group is the highest-level multiple profit center linking the corporate chief executive officer or
chief operating officer directly to two or more single profit center units (divisions).
A division is the lowest-level profit center responsibility for a business unit that engineers,
manufactures, and sells its own products.
A plant is a budget or cost center whose general manager supervises manufacturing, as well as
service functions, such as accounting, personnel, purchasing, and product engineering, but usually
no R&D engineering. More important, the manager of a plant never has sales responsibility.